The Position-sensorless Control of Low Voltage

High Power Permanent Magnet Synchronous

Motors in Zero/Low-speed Regions

Bo Guan, Shinji Doki Tomoyasu Furukawa, Norimoto Minoshima

Department of Electrical Engineering and TOYOTA INDUSTRIES CORPORATION
Computer Science, Nagoya University Aichi, Japan, 474-8601
Aichi, Japan 464-8603
Email: guanbo_1989@nagoya-u.jp

Abstract—The traditional maximum torque per ampere the field weakening control is required to achieve a higher
(MTPA) control and the position-sensorless control based on the speed driving. However, since the MTPA formulas and the
extended electromotive force (EEMF) with f-axis signal injection field weakening curve in [3][4] dependent on motor parameters
in zero/low-speed regions become difficult for low voltage high seriously, it just has theoretical significance when the motor
power permanent magnet synchronous motors (PMSMs), of parameters change dramatically.
which dq-axis inductances are susceptible to operating conditions.
Taking full account of the motor particularity, it is pointed out in A variable-step searching method of optimal current angle
this paper that the dramatic changes in dq-axis inductances will is raised in [5] to achieve the MTPA control of variable
lead to the incorrect orientation of maximum torque reference parameter Motors. Although it has a parameter robustness to
frame (f-t reference frame) and the failure of MTPA control. some extent, the searching process is very slow and the
Thereby, the EEMF will be inadequate for the position-sensorless dynamic performance of the motor control is bad. By injecting
control. For the disadvantages of the prior methods, a novel a high frequency signal to the current angle and comparing the
position-sensorless control of low voltage high power PMSMs in output torque, a gradual approaching method of optimal MTPA
zero/low-speed regions is proposed in this paper. By dealing with curve is presented in [6]. It overcomes the disadvantages of the
current control and position observation in different reference method in [5], which has a poor dynamic performance. The
frames separately, the MTPA control and the position-sensorless gradual approaching method needs the accurate information of
control are well realized. The simulation based on JMAG model
output torque, but generally there is no high accuracy torque
verifies the feasibility of the new proposal.
sensors in conventional control systems.
Keywords—low voltage high power; permanent magnet Recently position-sensorless control of PMSMs becomes a
synchronous motor; maximum torque per ampere; position- hot research. The position-sensorless control methods based on
sensorless control; incorrect orientation the electromotive force (EMF) observer [7] dominate in high-
speed regions, in which the rotor position is estimated by the
I. INTRODUCTION
inherent relationship of EMF and the rotor position. However,
Recently with the development of permanent magnetic the EMF decreases when the speed is low. It will be concealed
materials, PMSMs have been widely utilized in electric vehicle, by the system noise when the speed is less than 10% of rated
elevators, cranes, aerospace, etc [1]. PMSMs without excitation speed [8]. Thus, high-frequency signal injection methods are
current have a higher operating efficiency, especially under usually used to realize the position-sensorless control in zero/
light load [2]. High efficiency and high power density make low-speed regions. A high-frequency voltage signal is injected
PMSMs receive wide attention of scholars and some research intentionally to the stator and then the accurate rotor position
findings are got. can be obtained from the high-frequency component of the
This paper aims to discuss the position-sensorless control of stator currents. While this method can be used at less than 10%
low voltage high power PMSMs, of which the intrinsic of rated speed, the noise will increase and the performance of
parameters are very sensitive to operation conditions (dq-axis the output torque will deteriorate because of high frequency
inductances will strongly change up to 5 to 10 times). Due to voltage signal injection. [9] puts forward a position-sensorless
extremely few permanent magnet in the rotor, it can also be control method based on the EEMF observer with f-axis signal
seen as “quasi-reluctance motor”. Such special nature makes injection, which is applicable in zero/low-speed regions. By
common MTPA control and position-sensorless control be hard injecting a low-frequency current signal to f-axis current, which
to realize. For non-salient PMSMs, just by setting d-axis is insensitive to output torque, the position-sensorless control in
current to 0 and controlling q-axis current independently, zero/low-speed regions can be achieve based on the EEMF.
MTPA control can be realized. For salient PMSMs, since the This method is verified only for some motors, of which the
electromagnetic torque is a function of dq-axis currents, it is parameters change little. But for variable parameters motors, it
necessary to derive the MTPA curve or formula. At the same, can’t be used directly.

978-1-5090-3474-1/16/$31.00 ©2016 IEEE 2963

 Considering the problems of the prior methods for variable To eliminate the rotor position information in αβ-axis
parameters motors, based on detailed analysis and comparison, inductances, equation (1) can be rewritten as equation (2).
finally a novel position-sensorless control scheme is proposed
in this paper. éva ù é R + pLd wre ( Ld - Lq ) ù éia ù
êv ú = ê úê ú
II. POSITION-SENSORLESS CONTROL BASED ON THE EEMF ë û ë re ( d
b ê -w L - Lq) R + pL d
ú i
ûë b û (2)
A. The definitions of parameter and reference frames é - sin (q re ) ù
The definitions of reference frames are shown in Fig. 1. α-β {
+ ( Ld - Lq )(wreid - i&q ) + wre K E ê } ú
reference frame is a fixed reference frame. d-q reference frame ë cos (q re ) û
is a rotating reference frame. γ-δ reference frame is an The second term on the right side of equation (2) is defined
estimating reference frame. f-t reference frame is a maximum as extended electromotive force(EEMF), shown as equation (3).
torque reference frame. The parameters are defined as follows.
éea ù é - sin (q re )ù
[vα vβ]T αβ-axis voltages [iα iβ]T αβ-axis currents
êe ú = {( L }
- Lq )(wreid - i&q ) + wre K E ê ú (3)
ë cos (q re ) û
d
[vd vq]T dq-axis voltages [id iq]T dq-axis currents ë bû
[vf vt]T ft-axis voltages [if it]T ft-axis currents Thus, the position information is integrated into the EEMF,
which can be used to estimate the rotor position.
R stator resistance KE EMF coefficient
C. The position and velocity estimation based on the EEMF
Ld d-axis inductance Lq q-axis inductance
obserber
p differential operator β current angle In order to the rotor position, it is necessary to design an
EEMF observer. The state equations of PMSMs are shown as
equation (4).

éi ù éA A12 ù é i ù é B1 ù é0ù éi ù
p ê ú = ê 11 + v + ê ú ,i = C ê ú
ëeû ë 0 A22 úû êëe úû êë 0 úû W
ë û ëe û
ì æ R ö ì ( L - Lq ) üï
ï A11 = - ç ÷ I + ïíwre d ýJ
ï è Ld ø îï Ld
þï
ï
ï æ -1 ö
ï A12 = ç ÷ I = a12 I , A22 = wre J = a22 J (4)
ï è Ld ø
ï æ 1 ö
ï
í B1 = ç ÷ I , C = [ I 0]
ï è Ld ø
Fig.1 The definitions of reference frames ï
B. The extended electromotive force model of PMSMs
ïW = ( L - L ) w i& - i&& éê - sin (q re ) ùú
( )
ï ë cos (q re ) û
d q re d q

The mathematical model of PMSMs inα-β reference frame ï

is shown as equation (1). ï é1 0 ù é0 -1ù
ï I = ê 0 1 ú , J = ê1 0 ú
é va ù é R + pLa - pLab ù éia ù é - sin (q re ) ù ïî ë û ë û
êv ú = ê pL ú ê ú + wre K E ê ú
ë bû ë ab R + pLb û ëib û ë cos (q re ) û Thus, design the reduced order observer as equation (5).
ì L0 = ( Ld + Lq ) / 2, L1 = ( Ld - Lq ) / 2 (1)
ïï
ìi&ˆ = A i + A eˆ + B v
ï 11 12 1
í La = L0 + L1 cos ( 2q re ) , Lb = L0 - L1 cos ( 2q re )
ï
ïî Lab = L1 sin ( 2q re )
ï&
ï
&
( )
íeˆ = A22eˆ + G iˆ - i&
(5)

ïG = g1 I + g 2 J
î
The inducing voltage derived from the rotor rotation is
called EMF. From equation (1), it is easy to know that since d- To avoid the current differential, define an intermediate
axis inductance is equal to q-axis inductance (Ld=Lq) for non- variables ξ as equation (6).
salient motors, the rotor position information just exists in the
EMF. The position can be got based on the EMF observer. ïìx = eˆ + Gi (6)
However, since d-axis inductance is not equal to q-axis í& &
inductance (Ld≠Lq) for salient motors, the position information îïx = eˆ + Gi&
exists not only in the EMF but also in αβ-axis inductances. It is Equation (7) is obtained from equation (6) and (5).
difficult to obtain the rotor position information from the EMF.

2964
 ïìx& = ( GA12 + A22 ) x + GB1v + G ( A11 - GA22 - A22 ) i (7)
current resulting from the change of f-axis current generates q-
í axis current differential to maintain the EEMF. At the same
ïîeˆ = x - Gi time, changing f-axis current has little effect on the output
torque of the motor.
Essentially, the EEMF observer is equivalent to a low-pass
filter. By configuring the zero-pole point, the feature of the B. The EEMF method based on signal injection
EEMF observer can be adjusted. Finally, the rotor position is In f-t reference frame, t-axis current is utilized to control the
calculated as equation (8). output torque, and a sinusoidal signal is injected into f-axis
current, shown as equation (13).
æ eˆ ö
qˆre = tan -1 ç - a ÷÷ (8)
ìi f = I i sin (wi t )
ç eˆ (13)
è b ø í
î it = I t
From equation (3), it is easy to see that EEMF is directly
dependent on the rotor speed, EMF coefficient and q-axis Convert equation (13) into d-q reference frame.
current differential. Under steady-state conditions, q-axis
current differential is close to zero. Meanwhile the parameter ìïid = I i sin (wi t ) cos (j ) - I t sin (j )
í (14)
ïîiq = I i sin (wi t ) sin (j ) + It cos (j )
KE of the motors studied in this paper is very small. When the
rotor speed is below 10% of rated speed, the EEMF is
overshadowed by the system noise. Since It and φ change slowly with respect to ωit, they are
able to be seen as constants. Thus, q-axis current differential
III. THE POSITION-SENSORLESS CONTROL METHOD BASED ON
can be expressed as equation (15).
THE EEMF OBSERVER WITH F-AXIS DIGNAL INJECTION IN
ZERO/LOW-SPEED REGIONS i&q = wi I i cos (wi t ) sin (j ) (15)
Only q-axis current differential can be controlled to keep
the EEMF. However, q-axis current will affect the performance Equation (16) is obtained from equation (14) and (15).
of the output torque. Therefore, it is necessary to set q-axis
é - sin (q re ) ù
reference current with a special method, which will not e = { E0 - Ei sin (wi t - v )} ê ú
deteriorate the performance of the output torque. A position- ë cos (q re ) û
sensorless control based on the EEMF observer in f-t reference
ì
frame is proposed in [9].
{
ï E0 = wre K E + ( Lq - Ld ) I t sin (j ) } (16)
A. Maximum torque reference frame (f-t reference frame) ï
ï
í Ei = (wre cos (j ) ) + (wi sin (j ) ) ( Lq - Ld ) I i
2 2
The output torque (T) of PMSMs is equal to the outer
product of the current vectors (idq= [id, iq]T) and the flux vectors ï
(Ψo= [KE＋Ldid, Lqiq]T), as shown in equation (9). ï tan ( v ) = wi tan (j )
ïî wre
{
T = Pn KE - ( Lq - Ld ) id iq } (9)
Although E0 becomes very small in zero/low-speed regions,
Ei is still able to maintain the EEMF. Because the EEMF
The constant torque curve is a hyperbola, as shown in
observer has a low-pass filter function, it is very important for
equation (10).
ensuring the fundamental component and injected frequency
T component of the EEMF to configure the zero-pole point of the
iq = (10)
{
Pn K E - ( Lq - Ld ) id } observer reasonably. However, it makes the EEMF contain a
lot of noise. The pure EEMF can be obtained via synchronous
demodulation and a low pass filter, as shown in Fig. 2.
The tangent slope of the constant torque curve at any point
can be expressed as equation (11).

diq
=
(L q - Ld ) iq
(11)
did K E - ( Lq - Ld ) id

Thus, the angle between f-t reference frame and d-q

reference frame is defined as φ.

j = tan -1
(L q - Ld ) iq
(12)
KE - ( Lq - Ld ) id

On the one hand, whether the salient motor is, MTPA

control can be realized by setting f-axis current to 0 and
adjusting t-axis current. On the other hand, the change of q-axis Fig.2 The EEMF method based on signal injection

2965
IV. THE LIMITATIONS OF THE POSITION-SENSORLESS CONTROL A. The incorrect orientation of f-t reference frame in variable
BASED ON THE EEMF OBSERVER WITH F-AXIS DIGNAL parameter motors
INJECTION FOR VARIABLE PARAMETER MOTORS The accurate orientation of f-t reference frame is an
A low voltage high power PMSM is discussed in this paper. important precondition for the vector control. For variable
Its basic parameters is shown in Table I. Fig. 3 is the designed parameter motors, the following two factors may cause the
parameter map of JMAG model. orientation error. The tangent angle (φ) of the constant torque
curve can be calculated via equation (17).
TABLE I. THE PARAMETERS OF PMSMS
DC voltage 48V Lq/Ld 8 iq =
T
Þ
diq
=
(L q - Ld ) iq
Maximum Power 10kW Pole pairs
Switching frequency
2
8kHz
{
Pn K E - ( Lq - Ld ) id } did K E - ( Lq - Ld ) id (17)
EMF coefficient 0.00074Vrms/rpm
Stator resistance 4.6mΩ Dead time 5uS
Þ j = tan -1
(L q - Ld ) iq
K E - ( Lq - Ld ) id

On the one hand, if an error exists between the dq-axis

inductance used in equation (17) and the real dq-axis
inductance, the orientation error of f-t reference frame will
occur. The sensitivity of the orientation error for the error of
dq-axis inductances can be expressed as equation (18).
¶j K E iq
Dj L = DL = DL
¶L ( K E - Lid ) (18)
2

L = Lq - Ld , DL = DLq - DLd

(a) d-axis inductance On the other hand, dq-axis inductances are assumed as
constants by default when calculating equation (17). As can be
seen from Fig. 3, the real dq-axis inductances are the function
of dq-axis currents. And thus equation (17) could be rewritten
as equation (19).
T
iq = Þ
{ (
Pn K E - Lq ( id , iq ) - Ld ( id , iq ) id ) } (19)
diq
=
( L (i , i ) - L (i , i ) + L (i , i ) - L (i , i )) i
q d q d d q
'
q d q
'
d d q q

did K - ( L (i , i ) - L (i , i )) i
E q d q d d q d

Only when dq-axis inductances are the constants, equation

(b) q-axis inductance
(17) is equivalent to equation (19). Even if the accurate dq-axis
inductances can be get by looking up the inductance table, only
the error in equation (18) can be overcome. The correct
orientation angle of f-t reference frame cannot be calculated via
equation (19) since it is hard to obtain the accurate inductance
differential. The incorrect orientation of f-t reference frame will
seriously worsen the effect of the vector control.
B. The shortage of extended electromotive force derived from
the failure of MTPA control in f-t reference frame
Chapter 3.1 points out, the incorrect orientation will worsen
the vector control and thus increase the harmonic currents for
variable parameter motors. The further fluctuations of dq-axis
(c) the output torque
Fig.3 The designed parameter of JMAG model inductances seriously affect the performance of the EEMF
observer. It is easy to see from Fig. 3 that (Lq-Ld) can be
It is easy to know from Fig. 3 that dq-axis inductances of maintained at an adequate value in the right MTPA control and
the special motor change drastically with the operation points. may be too low to keep the EEMF in the wrong MTPA control.
Although the position-sensorless control based on the EEMF As is known from equation (3), the performance of the
observer with signal injection in zero/low-speed regions is position-sensorless control will be significantly deteriorated
feasible in theory, some problems still exist in the variable since the EEMF was seriously weakened.
parameter motor.

2966
V. THE POSITION-SENSORLESS CONTROL OF LOW VOLTAGE VI. SIMULATION
HIGH POWER PMSMS IN ZERO/LOW-SPEED REGIONS To verify the correctness of the analysis of the incorrect
Considering the incorrect orientation and the failure of orientation and the incorrect MTPA control and the feasibility
MTPA control in f-t reference frame, a feasible position- of the proposal method for the position-sensorless control of
sensorless control of low voltage high power PMSMs in low voltage high power PMSMs, the simulation is done based
zero/low-speed regions is proposed in this paper. To ensure the on JMAG model and Matlab/Simulink software.
EEMF, the correct MTPA control is an essential condition. As
The simulation results of the MTPA control in f-t reference
mentioned above, both the MTPA control in f-t reference frame
frame is shown in Fig. 6. 0-1s, the motor speed accelerates
and the robust MTPA methods mentioned in [3][4][5] are not
from 0 rpm to 300 rpm and the maximum torque is generated.
appropriate for variable parameter motors. Only the look-up
table method is the right choice. The relationship between the
current (Iam) and the current angle (β) under maximum output
torque can be obtained from Fig.3(c), as shown in Fig. 4

(a) Iu_Speed_Torque

Fig.4 The relationship between Iam and β

The fitting curve can be expressed as equation (20).

b = -31.5 I am
2
+ 61.03I am + 36.080658 (20)

In addition, due to the incorrect orientation of f-t reference

frame, the current control and position observation cannot be (b) dq-axis currents
achieved in f-t reference frame. This paper proposes a control
scheme dealing with current control and position observation
separately. In d-q reference frame, the look-up table method is
adopted to realize the correct MTPA control and field
weakening control. A sinusoidal signal given in f-t reference
frame is converted into d-q reference frame and then
superimposes onto the q-axis reference current, shown as Fig. 5.

(c) dq-axis inductances

Fig.6 The designed parameter of JMAG model

From Fig. 6, some results can be found: 1) the average

output torque is about 70% of rated torque, which is designed
in Fig. 3(c). And the output torque fluctuations intensely; 2) the
incorrect orientation of f-t reference frame cause a serious
distortion in dq-axis currents; 3) due to the incorrect MTPA
control and dq-axis harmonic currents, q-axis inductance will
be severely reduced and fluctuate dramatically. Accordingly,
Fig.5 The control diagram of low voltage high power PMSMs
under zero/low-speed regions
the position-sensorless control based on the EEMF observer
with f-axis signal injection becomes invalid for variable
On the one hand, the current control is implemented in d-q parameter motors. The simulation results of the position-
reference frame to avoid the incorrect orientation of f-t sensorless control of low voltage high power PMSMs in zero/
reference frame for variable parameter motors. On the other low-speed regions proposed in this paper is shown in Fig. 7.
hand, q-axis injection signal is essentially derived from f-t 0-1s, the motor speed accelerates from 0 rpm to 300 rpm and
reference frame, which will not affect the performance of the the maximum torque is generated. Starting from zero speed, the
output torque. position-sensorless control is utilized.

2967
 Under the correct MTPA control, the decay of q-axis
inductance becomes small in favor of maintaining the EEMF,
as shown in Fig 7(c). It can be found from Fig. 7(d) that even at
a very low speed, the EEMF is still sufficient for the position-
sensorless control due to the signal injection.
VII. CONCLUSION
For the special nature of the low voltage high power
PMSMs, a novel position-sensorless control in zero/low-speed
regions is discussed in this paper. The MTPA control and the
(a) Iu_Speed_Torque
position-sensorless control are realized well even for variable
parameter motors. The simulation based on JAMG model
confirms the defects of the traditional methods and verifies the
correctness of the proposed method. The following conclusions
can be drawn:
1) For variable parameter motors, the orientation problem
will appear in the position-sensorless control based on the
EEMF observer with f-axis signal injection.
2) The MTPA control fails due to the incorrect orientation
of f-t reference frame, which results in the shortage of the
(b) dq-axis currents
EEMF. Thus, the position-sensorless control will be affected.
3) Since a variety of robust MTPA control methods will fail,
only the look-up table method can be utilized for variable
parameter motors.
4) The proposed method are able to overcome the issues of
MTPA control and the orientation problem.
REFERENCES

(c) dq-axis inductances [1] G. Xie, K. Lu, D. S. Kumar and R. J. Riber, “High Bandwidth Zero
Voltage Injection Method for Sensorless Control of PMSM,” ICEMS,
Hangzhou, pp. 3546-3552, 2014.
[2] N. Bedetti, S. Calligaro and R. Petrella, “A Novel Approach to the
Design of Back-EMF Observer Based Sensorless Control of non-salient
PMSM Theoretical Analysis and Experimental Investigations,” ECCE,
Denver, pp. 3496-3503, 2013.
[3] S. Bolognani, R. Petrella, A. Prearo and L. Sgarbossa, “Automatic
Tracking of MTPA Trajectory in IPM Motor Drives Based on AC
Current Injection”, IEEE Transactions on Industry Applications, vol. 47,
no. 1, pp. 105-114, 2010.
[4] M. N. Uddin, T. S. Radwan and M. A. Rahman, “Performance of
(d) EEMF and position information Interior Permanent Magnet Motor Drive over Wide Speed Range”, IEEE
Transactions on Energy Conversion, vol. 17, no. 1, pp. 79-84, 2002.
[5] A. Dianov, Kim Young-Kwan, Lee Sang-Joon and Lee Sang-Taek,
“Robust self-tuning MTPA algorithm for IPMSM drives,” IECON,
Orlando, pp. 1355-1360, 2008.
[6] S. Kim, Y. D. Yoon, S. K. Sul and K. Ide, “Maximum Torque per
Ampere (MTPA) Control of an IPM Machine Based on Signal Injection
Considering Inductance Saturation”, IEEE Transactions on Power
Electronics, vol. 28, no. 1, pp. 488-497, 2012.
[7] Zhiqian Chen, M. Tomita, S. Doki and S. Okuma, “An extended
electromotive force model for sensorless control of interior permanent-
magnet synchronous motors”, IEEE Transactions on Industrial
Electronics, vol. 50, no. 2, pp. 288-295, 2003.
(e) position error [8] R. W. Hejny and R. D. Lorenz, “Evaluating the Practical Low-Speed
Fig.7 The result of the proposed control scheme Limits for Back-EMF Tracking-Based Sensorless Speed Control Using
Drive Stiffness as a Key Metric”, IEEE Transactions on Industry
It is easy to see from Fig. 7(a) that the output torque of the Applications, vol. 47, no. 3, pp. 1337-1343, 2011.
proposed method reaches the rated maximum torque and
[9] P. Saitoh, Y. Makaino and T. Ohnuma, “Adaptive Signal Injection
fluctuates small. Thus, the correct MTPA control is realized. Method Combined with EEMF-based Position Sensorless Control of
As is known from Fig. 7(b), eliminating the orientation IPMSM Drives”, IEEJ Journal of Industry Applications, vol. 4, no. 4, pp.
problem, the harmonic currents are significantly inhibited. 454-459, 2015.

2968

Powered by TCPDF (www.tcpdf.org)